Change Detection Based On Multi-temporal Remote Sensing Of Land Use

Obtaining the land use information timely, quickly and accurately, is the precondition and fundamental basis for governments at all levels to implement the Land and Resource planning, management, protection and rational utilization. Using multi-source, multi-temporal and multi-resolution remote sensing images to carry on the land use change detection, can obtain more objective, multiple periodic, multi-scale information of land changes. This paper researched the land use change detection with high resolution QuickBird images and multispectral Landsat data of the same detection area, and carried out the study from two aspects:First:Studied the commonly used methods of remote sensing change detection deeply. To address the limitation of traditional change detection methods and difficulties of using very high resolution images to detect changes, this paper used the features of the segment objects of QuickBird images to detect the change information, and obtained the final changes on feature detection information fusion. Then checked the result with a higher resolution remote sensing image and more accurate vector data.Besides, the study analyzed the applicability of the method in land use change detection.Second:Based on the need to know the land use changes on a certain time and we can’t access the information from historical vector data and high-resolution remote sensing images, used multispectral Landsat data to detect land use change information of the research area by the methods of data-dimension compression with some indices and supervised classification. The specific research contents and results of this paper are as follows:(1) Studied the basic theory of change detection from several aspects and summarized the principle of commonly used change detection methods. In addition, the research also analyzed the advantages and disadvantages of traditional change detection methods.(2) A village of the research area as an experimental object, used multi-temporal QuickBird images to detect the land changes. The study used the spectral features, texture features, shape features of the segment objects to do change detection, and then made fusion of the different change detection results to obtain the final result. For testing the experimental results, the paper used a higher resolution remote sensing image (0.2m aerial photo) and more accurate vector data to check the results. Experiments showed that the method can maximize detecting variation information, but it needed a large number of screening the change detection results to apply to the actual land use change detection project.(3) The whole study area as an experimental object, with the detailed analysis of the land use types of the research area, the research choosed three indices, Soil Adjusted Vegetation Index(SAVI),Normalized Difference Built-up Index(NDBI),Modified Normalized Difference Water Index(MNDWI), to represent three major land use classes, including vegetation, built-up land and water body. Then, reduced6multispectral bands of Landsat TM and7multispectral bands of Landsat ETM to three index bands generated from the original multispectral bands, and detected the land use changes by supervised classification results. Experimental results showed that the accuracy of compressed data classification images was higher than the original ones, which based on the same classification of samples and the same method. Moreover, used classification images of compressed data can effectively detect the change information.